# US 11,078,531 B2 · Deepsimulator method and system for mimicking nanopore sequencing

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,078,531 B2
- Document ID: US11078531B2
- Status: Granted
- Kind: B2
- Application number: 16/769,127
- Filed: 2018-10-30
- Published: 2020-11-26
- Granted: 2021-08-03
- Earliest claimed: 2017-12-13
- Assignees: King Abdullah University of Science and Technology
- Applicants: King Abdullah University of Science and Technology
- Inventors: Xin Gao; Yu Li; Sheng Wang; Renmin Han
- Web page: https://patentgrep.com/patent/US11078531B2

## CPC

- C12Q 1/6869: Methods for sequencing
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G16B 30/20: Sequence assembly
- G16B 40/10: Signal processing, e.g. from mass spectrometry [MS] or from PCR
- G16B 40/30: Unsupervised data analysis
- C12Q: MEASURING OR TESTING PROCESSES INVOLVING ENZYMES, NUCLEIC ACIDS OR MICROORGANISMS (immunoassay G01N33/53); COMPOSITIONS OR TEST PAPERS THEREFOR; PROCESSES OF PREPARING SUCH COMPOSITIONS; CONDITION-RESPONSIVE CONTROL IN MICROBIOLOGICAL OR ENZYMOLOGICAL PROCESSES
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16B: BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY

## Abstract

A method for sequencing biopolymers. The method includes selecting with a sequence generator module an input nucleotide sequence having plural k-mers; simulating with a deep learning simulator, actual electrical current signals corresponding to the input nucleotide sequence; identifying reads that correspond to the actual electrical current signals; and displaying the reads. The deep learning simulator includes a context-dependent deep learning model that takes into consideration a position of a k-mer of the plural k-mers on the input nucleotide sequence when calculating a corresponding actual electrical current.

## Claims

### Claim 1

A method for sequencing biopolymers, the method comprising:
selecting with a sequence generator module an input nucleotide sequence having plural k-mers;
simulating with a deep learning simulator, actual electrical current signals corresponding to the input nucleotide sequence;
identifying reads that correspond to the actual electrical current signals; and
displaying the reads,
wherein the deep learning simulator includes a context-dependent deep learning model that takes into consideration a position of a k-mer of the plural k-mers on the input nucleotide sequence when calculating a corresponding actual electrical current, and
wherein the context-dependent deep learning model calculates transformed signals by using a Bi-LSTM extended Deep Canonical Time Warping, which combines a bi-directional long short-term memory (Bi-LSTM) method with a deep canonical time warping (DCTW) method.

### Claim 2 (depends on 1)

The method of claim 1, wherein the context-dependent deep learning model compares two linearly structured data sets having different lengths, and feature dimensionality, wherein the first data set corresponds to the input nucleotide sequence and the second data set corresponds to measured electrical current signals.

### Claim 3 (depends on 2)

The method of claim 2, further comprising:
applying a first deep neural network algorithm to the input nucleotide sequence to obtain the first data set; and
applying a second deep neural network algorithm to the measured electrical current signals to obtain the second data set,
wherein the first and second data sets are in a common space.

### Claim 4 (depends on 3)

The method of claim 3, further comprising:
applying an objective function to the first and second data sets to temporally align the input nucleotide sequence and the measured electrical current signals.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
repeating the transformed signals, in a signal repeating module, at each position based on a mixture of alpha distributions, to generate the actual electrical current signals.

### Claim 6 (depends on 5)

The method of claim 5, further comprising:
adding a random noise to the transformed signals.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
using plural different k-mers for each base of the input nucleotide sequence.

### Claim 8 (depends on 7)

The method of claim 7, wherein the plural different k-mers include a 1-mer, a 3-mer and a 5-mer.

### Claim 9 (depends on 7)

The method of claim 7, further comprising:
encoding the bases of the input nucleotide sequence with a one-hot encoding using the plural different k-mers for each base.

### Claim 10 (depends on 1)

The method of claim 1, wherein the step of selecting further comprises:
randomly selecting a starting position along the input nucleotide.

### Claim 11 (depends on 10)

The method of claim 10, further comprising:
selecting a length of a read based on one of three distributions.

### Claim 12

A computing device for sequencing biopolymers, the computing device comprising:
a processor for,
selecting with a sequence generator module an input nucleotide sequence having plural k-mers;
simulating with a deep learning simulator, actual electrical current signals corresponding to the input nucleotide sequence; and
identifying reads that correspond to the actual electrical current signals; and
a display for displaying the reads,
wherein the deep learning simulator includes a context-dependent deep learning model that takes into consideration a position of a k-mer of the plural k-mers on the input nucleotide sequence when calculating a corresponding actual electrical current, and
wherein the context-dependent deep learning model calculates transformed signals by using a Bi-LSTM extended Deep Canonical Time Warping, which combines a bi-directional long short-term memory (Bi-LSTM) method with a deep canonical time warping (DCTW) method.

### Claim 13 (depends on 12)

The device of claim 12, wherein the context-dependent deep learning model compares two linearly structured data sets having different lengths, and feature dimensionality, wherein the first data set corresponds to the input nucleotide sequence and the second data set corresponds to measured electrical current signals.

### Claim 14 (depends on 13)

The device of claim 13, wherein the processor is further configured to:
apply a first deep neural network algorithm to the input nucleotide sequence to obtain the first data set; and
apply a second deep neural network algorithm to the measured electrical current signals to obtain the second data set,
wherein the first and second data sets are in a common space.

### Claim 15 (depends on 14)

The device of claim 14, wherein the processor is further configured to:
apply an objective function to the first and second data sets to temporally align the input nucleotide sequence and the measured electrical current signals.

### Claim 16 (depends on 12)

The device of claim 12, wherein the processor is further configured to:
repeat the transformed signals, in a signal repeating module, at each position based on a mixture of alpha distributions, to generate the actual electrical current signals.

### Claim 17 (depends on 16)

The device of claim 16, wherein the processor is further configured to:
add a random noise to the transformed signals.

### Claim 18

A non-transitory computer readable medium including computer executable instructions, wherein the instructions, when executed by a processor, implement instructions for sequencing biopolymers, the instructions comprising:
selecting with a sequence generator module an input nucleotide sequence having plural k-mers;
simulating with a deep learning simulator, actual electrical current signals corresponding to the input nucleotide sequence;
identifying reads that correspond to the actual electrical current signals; and
instructing a display to display the reads,
wherein the deep learning simulator includes a context-dependent deep learning model that takes into consideration a position of a k-mer of the plural k-mers on the input nucleotide sequence when calculating a corresponding actual electrical current, and
wherein the context-dependent deep learning model calculates transformed signals by using a Bi-LSTM extended Deep Canonical Time Warping, which combines a bi-directional long short-term memory (Bi-LSTM) method with a deep canonical time warping (DCTW) method.